DE69006518T2 - Semiconductor laser amplifier. - Google Patents

Abstract

A semiconductor laser amplifier includes both a waveguide layer (3) and an active layer (4). In use, an optical wave in the active layer (4) interacts with the waveguiding layer to reduce the polarisation sensitivity of the amplifier. A short, thick active layer is not required, typical dimensions being 250 or 500 mu m long, by 0.2 mu m high. The mesa (15) is typically about 2 mu m wide, but may be wider.

Description

The present invention relates to semiconductor laser amplifiers.

Semiconductor laser optical amplifiers have many potential applications in future optical systems, particularly in optical communication systems and optical data processing. Such amplifiers provide high gain at low power consumption and their single-mode waveguide structure makes them particularly suitable for use with single-mode optical fibers.

The radiation used in optical communication is not necessarily in the visible range and the words "optical" and "light" when used in this specification should not be interpreted as implying any such limitation. In particular, when silica optical fibers are used as the transmission medium, infrared radiation is of particular utility since loss minima in such fibers occur at about 1.3 and 1.55 µm.

In optical transmission systems using single-mode longitudinal lasers, the effects of fiber dispersion can be small, the main limitation on the spacing of the line amplifiers being signal attenuation due to fiber loss. Such systems do not require complete regeneration of the signal at each line amplifier, amplification of the signal is sufficient. Accordingly, semiconductor laser amplifiers can be used as linear optical line amplifiers for intensity-modulated or coherent systems. Linear line amplifiers, and in particular traveling-wave line amplifiers, have the additional advantages of being bitrate independent, bidirectional and able to support multiplexing.

A semiconductor laser amplifier can be used in an optical receiver to amplify the optical signal before it reaches the photodetector, thereby increasing detection sensitivity. The improvement can be particularly noticed for bit rates greater than 1 Gbit/s, making such amplifiers potentially useful in developing sensitive broadband receivers for use in future fiber systems.

The performance of laser amplifiers in systems is dependent on several factors, including amplifier gain and saturation characteristics, optical amplifier bandwidth, stability of the gain with respect to bias current and temperature, amplifier noise characteristics, and sensitivity to input signal polarization. Clearly, for any practical application it is desirable that the gain be frequency independent, at least within the frequency range of the optical input. This is of course of particular importance if the laser amplifier is to be used in systems involving wavelength division multiplexing (WDM), or in systems where optical carriers of several different wavelengths are used, particularly where several such carriers are used simultaneously. A second desirable characteristic for a practical system is that the gain of the laser amplifier should be independent of the input signal polarization. This last characteristic is of particular importance in optical fiber systems, since random polarization variations are currently unavoidable in practical systems. Much research has been done with the purpose of providing high gain laser structures over a useful range of optical bandwidth with suitably low polarization sensitivity. Accordingly, although usefully high gains have been achieved typically between 24 and 29 dB, polarization sensitivity remains a problem. In particular, although polarization sensitivities as low as 2 to 3 dB have been achieved, typical sensitivities are between 3 and 6 dB and the results were strongly frequency dependent, polarization sensitivities of 7 dB or more are typical at non-preferred frequencies, thus laser amplifiers cannot be used satisfactorily in WDM systems.

In recent demonstrations of traveling wave semiconductor laser amplifier (TWSLA) systems (see, e.g., Proc. ECOC, Brighton, 1988, pages 163 to 166, D. J. Malyon et al, and Electronic Letters, Vol. 24, 1988, pages 551 to 552, G. Grosskopf et al), the polarization of the input signal had to be carefully controlled due to polarization dependent gain. The problem of polarization dependent gain appears to be exacerbated in device designs that yield highly saturated output power. In fact, the TWSLA with the highest reported output at 3 dB gain compression (+9 dBm) had a polarization sensitivity of 7 dB, Electronic Letters, Vol. 23, 1987, pages 218 to 219, T. Saitoh and T. Mukai. Gain variations are possible either through electronic devices (Proceedings, ECOC, Brighton, 1988, pages 487 to 490, A. D. Ellis et al) or by using orthogonally aligned two-amplifier configurations (Electronic Letters, Volume 24, pages 1075 to 1076, N. A. Olsson; Electronic Letters, Volume 23, 1987, pages 1387 to 1388, G. Grosskopf et al). Although these schemes have been successfully demonstrated, additional components are required for their implementation.

It would clearly be desirable to have a TWSLA with low polarization sensitivity, but not at the expense of the high saturated power required for a wide dynamic range and good signal-to-noise efficiency (see e.g. IEEE Trans, MTT-30, 1982, pages 1548 to 1556, T. Mukai et al).

In a recent theoretical paper (Transaction of the IEICE, Vol. E71, No. 5, May 1988, pages 482 to 484), Saitoh and Mukai reported that the optimal configuration for achieving polarization-insensitive gain in a traveling-wave semiconductor laser amplifier, while maintaining high output power and low noise efficiency, is one with a thick, short active layer. Saitoh and Mukai predict that an undoped 1.5 µm GaInAsP traveling-wave amplifier with an active layer thickness of 0.5 µm and an active layer length of 45 µm could have a 20 dB signal gain at an operating current of 100 mA and a polarization sensitivity of only 0.6 dB. They also predict that under these conditions such a device would have a saturated output power P3dB of 9.6 dBm and a noise figure (the deterioration in the S/N ratio resulting from the gain) of 4 dB.

One of the disadvantages of structures of the type proposed by Saitoh and Mukai is that with such a thick active layer there is a high probability that wave modes other than the zero-order wave mode will be supported in the vertical direction. In addition, the high current densities resulting from the use of short devices are likely to lead to reliability and thermal dissipation problems.

It has previously been proposed to use laser amplifiers with a waveguide layer separate from the active layer, but generally this has been in the context of optically integrated circuits (OIC). In such circuits, a waveguide arrangement is used to pipe the optical signals to be processed around the circuit; the active layer of a laser amplifier is spaced from the waveguide and operates in conjunction with it in a directional coupler mode. Such an arrangement is described in outline in British Patent 1539028. The waveguide layer in such arrangements appears to be intended only to act as a conduit to conduct optical signals between the OIC input and output and various processing points in the OIC, the optical signals having either initially been provided by a laser source forming part of the OIC or having been received by the OIC from a remote source. To our knowledge, no one has observed or suggested that the effective addition of a waveguide layer to a laser amplifier structure results or is likely to result in a reduction in polarization sensitivity.

In another reference related to OIC, IEEE Journal of Quantum Electronics, QE-23, 1987, No. 6, pages 1021 to 1026, Oberg, Bröberg and Lindgren, a laser amplifier is integrated in a passive waveguide structure. The aim of the reported work was to establish, and if possible demonstrate, the possibility of fabricating such a composite structure in which the connection between active and passive parts gives a high coupling efficiency and a reflectivity low enough to allow a travelling wave type operation of the laser amplifier. Again, the passive waveguide structure serves as a conduit for the distribution of optical signals to be processed. It is stated in the paper suggested that such OICs could be used in local area network switching systems instead of discrete amplifiers, with the OICs providing the amplification and other necessary functions, thereby eliminating interconnect difficulties, coupling losses and unwanted reflections. Of course, traveling wave operation for laser amplifiers is desirable in such applications because of the need to minimize wavelength sensitivity.

The structure used by Oberg et al is unusual in that the passive waveguide and amplifier active layer are formed on a common planar n-InP buffer layer. A p-InP "depression" layer is formed above the active InGaAsP layer (λg = 1.3 µm). The p-InGaAsP passive waveguide (λg = 1.08 µm) has a thickness of 0.55 and 0.3 µm near and far from the junction, respectively. The passive waveguide material continues across the depression layer which overlies the active layer, with the total thickness of the waveguide-depression layer-active layer stack apparently being the same as the immediately adjacent region of passive waveguide material, i.e. approximately 0.55 µm. The active and depression layers are 0.22 and 0.18 µm thick, respectively, or both 0.15 µm thick. It therefore appears that in the active region the waveguide-depression-active layer stack functions as a composite waveguide and not in a directional coupler mode. However, there is no suggestion that such a structure should be fabricated for use as a discrete device, the waveguide is clearly only envisioned for use as a signal line in an OIC. Oberg et al also, not surprisingly, say nothing about the polarization sensitivity of an amplifier with such a structure.

However, we are aware of a discrete laser amplifier having an active layer and a separate waveguiding layer, described in US Patent 4,742,307 in the name of Thylen, assigned to L.M. Ericsson. The device is intended to overcome the problems of back reflection and rear-radiated noise encountered in conventional amplifiers. To overcome these problems, the device operates on the principle of a directional coupler. Light to be amplified is input to a passive waveguide which continues to the output facet where it is aligned with an output fiber. Spaced from the passive waveguide in a directional coupler relationship is the active layer of the device. The ends of the active layer of the device are left unpumped or are etched to provide a rough surface, the aim being to reduce the amount of rearward traveling signals coupled back into the passive waveguide. Because the device must operate in a directional coupler mode, the separation between the active and waveguiding layers must be fairly wide, e.g. 1 µm. There is no suggestion anywhere that the device has a low polarization sensitivity or is in any way suitable for receiving input signals that are variable in polarization state.

The present invention seeks to provide a semiconductor laser amplifier that has high gain and low polarization sensitivity over a usefully wide frequency range.

According to a first aspect, the present invention provides a semiconductor laser amplifier having no grating, the amplifier comprising an active layer separated by a first epitaxial layer, a waveguide layer separated from the active layer and provided by a second epitaxial layer, and having an input facet and an output facet, the waveguide layer and the active layer each extending to both of the facets and being separated from each other by a distance less than 1 micron, the arrangement being such that, in use, an optical wave in the active layer interacts with the waveguide layer, thereby reducing the polarization sensitivity of the amplifier.

We have discovered the surprising fact that by using a structure analogous to a distributed feedback laser and a buried heterostructure, having a waveguide layer and an active layer in close association, but without the grating, it is possible to achieve high gain and low polarization sensitivity over a usefully wide range of wavelengths.

Embodiments of the invention will now be described by way of example only with reference to the accompanying drawings, in which:

Figure 1 is a schematic perspective view of an embodiment of a laser amplifier according to the present invention ;

Figure 1a is a schematic vertical section along the line A-A through the device of Figure 1;

Figure 2 is a similar schematic vertical section through a distributed feedback laser similar in structure to the amplifier of Figure 1;

Figure 3 is a graph showing the gain of an optical amplifier according to the invention for orthogonally input polarization states;

Figure 4 is a graph illustrating the gain saturation characteristics of a laser amplifier according to the invention measured at 1516 nm;

Figure 5 is a schematic diagram of the experimental setup used in producing the results reproduced in Figures 3 and 4; and

Figure 6 shows a series of photoluminescence curves for a series of devices in which the separation between the waveguide layer and the active layer was varied.

Referring now to Fig. 2, a distributed feedback (DFB) laser of known type comprises a buried heterostructure mesa in which the grating 2 providing the feedback necessary for laser operation is grown in a surface of a semiconductor layer 3 other than the active layer 4 of the laser, the former layer functioning as a waveguide layer, and in addition a spacer layer 5 is provided separating the waveguide layer and the active layer. The spacer layer 5 is provided for several reasons:

(i) Typically, the waveguide layer 3 and the active layer 4 must have different refractive indices, but both can be quaternary semiconductors based on the same elements, although of course they have different compositions.

During the growth of the heterostructure, abrupt changes in gas flow are expected if layer 3 were grown immediately after layer 4 or vice versa. By growing a spacer layer with a completely different composition, i.e. a binary rather than a quaternary semiconductor, on top of the first growth of layers 3 and 4, it is easily possible to keep the gas flow rate fairly constant and still produce the desired properties in layers 3 and 4. As one skilled in the art of heterostructure fabrication will appreciate, while the spacer layer, which in the present example comprises InP, is grown using, for example, gas flow A, gas flow B which has just been used to grow the InGaAsP active layer will be stabilized for the InGaAsP waveguide layer composition and will go to waste. When the spacer layer is complete, a gas stream A will be sent to waste and a gas stream B will enter the reactor.

(ii) The spacer layer also facilitates etching of the mesa structure using selected wet chemical etchants in the composition.

iii) Of more direct relevance in the context of distributed feedback lasers, it is possible to control the optical coupling to the grating in the waveguide layer by adjusting the thickness of the spacer layer. The spacer layer should also suppress carrier overflow.

The laser amplifier of Fig. 1 was fabricated using three stages of MOVPE growth, further details of which can be found in the following two publications: Nelson AW, Devlin, WJ, Hobbs, RE, Lenton, CGD and Wong, S., "High power, low threshold BH lasers operating at 1.52 um grown entirely by low pressure MOVPE", Elec. Lett., 1985, vol. 21, pages 888 to 889; and Cooper, DM, Evans, JS, Lealman, IF, Regnault, JC, Spurdens, PC and Westbrook, LD, "Properties of detuned BH distributed feedback lasers grown by MOVPE", paper ThK2 OFC '88, New Orleans, Louisiana, USA.

All three stages are carried out in a horizontal reactor operating at atmospheric pressure using the metal alkyls trimethyl indium (TMIn) and trimethyl gallium triethyl phosphorus (TMGa.TEP) as the Group III sources and arsine and phosphine as the Group V element precursors. Dimethyl zinc and hydrogen sulfide provided the n- and p-dopings, respectively, and all growth stages were carried out at 650 ºC. The first stage involved growing a planar structure on an InP wafer 1 consisting of an n-type InP buffer layer 6.2 µm thick, followed by a 0.18 µm thick InGaAsP active layer 4 (λ = 1.53 µm), a p-type InP spacer layer 5 0.16 µm thick, an InGaAsP waveguide layer 3 (λ = 1.13 µm) 0.23 µm thick and finally a thin (0.05 µm) p-type InP cap layer 7.

A mesa was then formed using a SiO2 mask, conventional photolithography, and selected etchings of HBr and 3(IN):1:1 (K2Cr2O7:H2SO4:HCl). The width of the waveguide layer was 2.0 µm and the width of the active layer after grooving was 1.4 µm.

Current blocking layers 8 of p- and n-InP were selectively grown around the mesa. After removing the oxide, the entire device was covered with a p-type InP layer and a p+ contact layer 10 of InGaAs. Contacts were made using sputtered Ti/Pt/Au 11 on the p-side and Ti/Au 12 on the n-side.

Typical device thresholds for 250 µm and 500 µm devices before chopping were approximately 12 mA and 9 mA, respectively. Individual laser amplifiers were formed by chopping the wafers into bars, 500 µm wide, which were then facet coated with a single layer coating of HfO₂ (the devices had a fundamental facet reflectivity of approximately 0.25%) or with broadband, three-layer coatings of (i) alumina, silicon and magnesium fluoride or (ii) alumina, silicon and silicon dioxide (given a fundamental facet reflectivity in the range of 0.01 to 0.1%). The threshold current of the device after AR coating was typically 100 mA. Finally, the laser amplifiers were mounted p-side down on diamond heat sinks using Au/Sn solder.

Fig. 5 shows the experimental setup used to establish the performance of the TWSLA according to the invention. The laser amplifier 20 being tested receives light from a light source 22 via a fiber 21 terminated with an AR coated lens. Between the light source 22 and the laser amplifier 20 there is an attenuator 23, a 3 dB coupler 24 and a polarization controller 25. The coupler 24 allows the optical input power to be monitored. The facet input power to the amplifier was -28 dBm. The output of the laser amplifier 20 is fed into a further Fiber 26 is coupled to the lens end and then via another 3 dB coupler 27 and an attenuator 28 into an optical receiver 29.

DFB lasers with a buried heterostructure (BH), also fabricated by the full MOVPE process (the above-mentioned paper ThK2 by Cooper, D.M. et al), were used as the optical source 22. These DFB lasers were temperature tunable, each with a tuning range of 3 nm. Lasers with wavelength ranges centered at 1503, 1509 and 1516 nm were used in the experiments.

Fig. 3 shows the internal TE and TM gain of an example of a TWSLA according to the invention as a function of bias current. The effects of fundamental Fabry-Perot resonances were allowed for by tuning the different polarizations. Typically there is a 3 dB passband ripple at a bias current of 95 mA.

As can be seen from the figure, the TWSLA described here shows a very low polarization sensitivity. In fact, the polarization sensitivity of the device was less than the measurement uncertainty of 1 dB (the measurement uncertainty was estimated by repeated measurements under the same conditions and was found to be similar to the scatter in the overall data), with an associated gain of 30 dB. The measurements of Fig. 3 are for a source wavelength of 1509 nm. Similar results were obtained with the same device at 1503 and 1516 nm.

Fig. 4 shows the results of measuring the gain of the device as a function of the input power, where the device is designed for unsaturated Gain of 20 and 28 dB: the saturated output power was 0 and 5.2 dB respectively at the compression point of 3 dB.

In devices according to the invention, the polarization dependence of the optical confinement factor has been largely eliminated.

As previously mentioned, the laser amplifier shown in Fig. 1 has been developed from a distributed feedback laser. Since the laser amplifier is made without a grating, the main reasons for forming the waveguide layer after the active layer (these are to avoid "melting back" of the grating and the need to grow the delicate active layer of the device above the grating) no longer exist.

Accordingly, it is feasible to grow the waveguide layer before the active layer. When this is done, the spacer layer is no longer needed to facilitate etching of the mesa structure using compositionally selected wet chemical etchants, although it should be noted that since the Q1.5 active layer etches faster than the Q1.1 waveguide layer in the present mesa definition process, this would likely lead to problems in etching the complete mesa as described above because of the waveguide layer beneath the active layer. An alternative structure which avoids this and other problems of simple inversion is now described. Similarly, the spacer layer is no longer needed as a means of controlling the optical coupling to the grating. Therefore, the spacer layer in such an "inverted" structure would serve only one of its original purposes, namely, maintaining a stable gas flow during the various growth stages. We have found that a stable gas flow can be maintained without having to grow a spacer layer, in which case the spacer layer can be omitted. Of course, in such inverted structures the waveguide would be n-type rather than p-type and, since the carrier mobility and recombination will be different, this may lead to different optical efficiency.

The "alternative" structure referred to above is based on the following planar. On an n-type InP substrate, an n-InP buffer layer of typically 2 µm thickness, a GaInAsP waveguide (λ = 1.1 µm) of 0.2 µm thickness, an n-InP spacer of 0.2 µm thickness, an undoped GaInAsP active layer (λ = 1.5 µm) of 0.2 µm thickness and a p-InP cap of 0.3 µm thickness are grown. A mesa is formed as normal by selectively etching the p-InP, Q1.5 (active layer) and n-InP layers in turn. The Q1.1 (waveguide) layer is not etched and this is expected to have advantages in terms of reliability. The potential disadvantages are the need to grow on a Q1.1 surface and the limited height available to fit into blocking layers. The following advantages are believed to more than compensate for these potential disadvantages: the process is simpler than the first described, precise control of the mesa height is possible, the electronic properties of the blocking layers are likely to benefit from reduced hole injection - holes trapped by the valence band discontinuity in the double Q1.1/InP heterostructure.

Another useful modification of the structure described above is to use a mesa that is wider than normally found in a Laser oscillator is allowed. Increasing the width is possible in non-resonant, low reflectivity devices such as TWSLA because the optical wave only makes one main pass through the device. Typically, the optical input to the laser amplifier is provided by a single-mode optical fiber. The fundamental mode will be coupled through the input fiber and there may not be enough spacing to couple much power into the higher order modes allowed by a wider structure. One advantage of using a wider structure is that increased output power may result.

As well as fabricating devices with waveguide and active layers separated by a 0.2 µm spacer layer, other separations have been attempted. In Fig. 6, photoluminescence curves are shown for four device types, each device type having a different waveguide/active region separation. The photoluminescence curves were all fabricated using devices with active layers having a quaternary composition (GaInAsP) and having a wavelength equivalent to 1.3 µm, and with waveguide layers having a quaternary composition (GaInAsP) having a wavelength equivalent to 1.1 µm. For the four curves, the separation was 0.2 µm, 0.11 µm, 0.04 µm, and 0. The photoluminescence curves illustrate that even for a separation as small as 0.04 µm, there is recombination in the waveguide layer different from that in the active layer. As the separation between the active and waveguide layers increases, the fraction of recombination occurring in the waveguide (at least during the photoluminescence measurement) increases at the expense of recombination in the active layer. In all our device configurations, the active layer and the waveguide are optically very tightly coupled. We believe this is important if low polarization sensitivity is to be achieved for the waveguide layer and the active layer to act effectively as a single waveguide - i.e. without any directional coupler characteristics. With only a small distance, of about 0.2 µm, between the waveguiding and active layers, it is possible to feed the optical output from a lensed fiber simultaneously into the active and waveguiding layers. In the vast majority of our experiments, such an arrangement has been used with good results. We have also, as far as possible with a lensed fiber end, directed the optical input signal primarily to the active region with spillover onto the waveguide layer. Such an arrangement also gave low polarization sensitivity and it is believed that this approach is preferred when there is a large separation between the active and waveguiding layers. To obtain low polarization sensitivity, we believe that the separation between the active and waveguiding layers should be considerably less than 1 µm. Preferably, we believe that the separation should not be more than 0.7 µm, and even more preferably it should not be more than 0.6 µm. As one skilled in the art will appreciate, the precise upper limit in each case will depend on the refractive indices of the active, waveguiding and separating layers, since these factors all affect the degree of coupling between the active and waveguiding layers.

Although in the examples described above the waveguide layers were shown comprising quaternary material with a wavelength equivalent to 1.1 µm, it is of course possible to use other compositions provided that they have a larger band gap than that of the active layer. For example, in Devices having active layers with a composition having a wavelength equivalent to 1.5 µm have successfully used waveguide layers with a wavelength equivalent to 1.3 µm.

Claims (20)

1. A semiconductor laser amplifier not comprising a grating, the amplifier comprising an active layer defined by a first epitaxial layer, a waveguide layer separated from the active layer and provided by a second epitaxial layer, and an input facet and an output facet, the waveguide layer and the active layer each extending to both of the facets and being separated from each other by a distance less than 1 micron, the arrangement being such that in use an optical wave in the active layer interacts with the waveguide layer, thereby reducing the polarization sensitivity of the amplifier. 2. An amplifier according to claim 1, wherein the separation between the active layer and the waveguide layer is small enough so that in use, significant exchange of modal power between the layers is avoided. 3. An amplifier according to claim 1 or claim 2, wherein the active layer is grown over the waveguide layer. 4. Amplifier according to claim 1 or 2, wherein the waveguide layer is grown over the active layer. 5. An amplifier according to any preceding claim, wherein the separation between the active layer and the waveguide layer is 0.6 microns or less. 6. An amplifier according to any preceding claim, wherein the active layer comprises a quaternary semiconductor having a band edge equivalent wavelength between 1.3 and 1.55 microns. 7. An amplifier according to any preceding claim, wherein the active layer has a thickness of less than 0.4 µm. 8. An amplifier according to claim 7, wherein the active layer has a thickness of at least 0.15 microns. 9. An amplifier according to any preceding claim, wherein the active layer has a width of not more than 2 µm. 10. An amplifier according to any preceding claim, wherein the active layer has a length of 250 µm or more. 11. Amplifier according to claim 10, wherein the active layer has a length of more than 300 µm. 12. An amplifier according to claim 10 or 11, wherein the active layer has a length of 500 µm or less. 13. An amplifier according to any preceding claim, wherein the reflectivities of the end facets of the amplifier are not more than 0.25% at the operating wavelength of the amplifier. 14. Amplifier according to one of the preceding claims, wherein the polarization sensitivity of the amplifier does not exceed 3 dB over the frequency range 1500 to 1520 nm. 15. Amplifier according to claim 14, wherein the polarization sensitivity does not exceed 1 dB over the frequency range 1500 to 1520 nm. 16. Amplifier according to one of the preceding claims, wherein the saturated output power is at least 5 dBm. 17. An arrangement comprising a semiconductor laser amplifier according to any one of the preceding claims and an input device arranged, in use, to direct optical input signals having a variable polarization state to the input facet of the laser amplifier, the input device, in use, causing the optical input signals to impinge on both the active and waveguide layers of the input facet. 18. An arrangement according to claim 17, wherein in use the input means is arranged so that the input optical signals at the input facet provide a greater optical intensity on the active layer than on the waveguide layer. 19. An arrangement according to claim 17 or 18, wherein the input means comprises an optical fiber and a lens means. 20. A method of amplifying variable optical polarization signals, the method comprising supplying the optical signals to be amplified to a facet of a semiconductor laser amplifier, the laser amplifier comprising an active layer and a separate waveguide layer each extending as far as the facet and separated by no more than 1 µm, and the amplifying optical input signals impinge on both the waveguide layer and the active layer of the facet.